AI Job Application Copilot
Human-in-the-loop resume tailoring grounded in verified evidence
Monorepo with a Next.js workspace, a FastAPI backend and a Chrome extension that score job fit, propose bullet-level resume changes for approval, verify missing skills, and turn gaps into portfolio projects.
Visual evidence
Resume studio with change inspector
Supplementary evidence
Explainable fit breakdown
Case study
Problem, Build, Outcome
Problem
Resume tools either keyword-stuff or rewrite claims the candidate cannot defend, and they never ask for the evidence that would make a stronger case.
Build
Built an experience vault of evidence-backed claims, a match engine that scores fit across skills, seniority and domain, a diff studio where every suggested change carries a reason, source and confidence, and PDF/DOCX export.
Outcome
Candidates approve or reject each change with its evidence visible, and the extension brings the same match scoring to LinkedIn, Indeed, Greenhouse and Lever postings.
Architecture
System Flow
Architecture Notes
- Shared TypeScript contracts keep the web app and the extension in sync.
- 23-table schema with Alembic migrations for PostgreSQL or SQLite.
- FastAPI services for matching, tailoring, outreach drafting and project generation.
Engineering choices
Technical Decisions
Required explicit approval for every resume change instead of auto-applying edits.
Grounded rewrites in stored evidence so claims stay defensible in interviews.
Made the shell responsive with a mobile drawer so the workspace works on a phone.
Evidence
Proof Points
Bullet-level change review with reason, requirement addressed and evidence source
Skill verification that asks before assuming a gap
Manifest V3 extension sharing typed contracts with the web app
Next read